Papers by Dennis Wei
Multi-component Causal Tracing in Large Language Models (2026.acl-long)
Copied to clipboard
| Challenge: | Large language models (LLMs) are prone to various forms of safety risks, such as learning and propagating societal biases and even creating harmful or deceptive content through jailbreak attacks. |
| Approach: | They propose a framework for causally tracing multiple components simultaneously that systematically identifies the subsets of components most critical to a desired performance metric. |
| Outcome: | The proposed method outperforms existing methods in identifying components critical to a desired performance metric. |
Reasoning Model Unlearning: Forgetting Traces, Not Just Answers, While Preserving Reasoning Skills (2025.emnlp-main)
Copied to clipboard
Changsheng Wang, Chongyu Fan, Yihua Zhang, Jinghan Jia, Dennis Wei, Parikshit Ram, Nathalie Baracaldo, Sijia Liu
| Challenge: | Existing methods for LRM unlearning overlook critical information leakage in reasoning traces, even when final answers are successfully removed. |
| Approach: | They propose a method that suppresses reasoning traces while preserving the model's general reasoning ability. |
| Outcome: | The proposed method significantly reduces reasoning trace leakage and achieves strong performance across reasoning and safety benchmarks, including WMDP, StrongReject, JBB-Behaviors and WildJailbreak. |
Multi-Level Explanations for Generative Language Models (2025.acl-long)
Copied to clipboard
Lucas Monteiro Paes, Dennis Wei, Hyo Jin Do, Hendrik Strobelt, Ronny Luss, Amit Dhurandhar, Manish Nagireddy, Karthikeyan Natesan Ramamurthy, Prasanna Sattigeri, Werner Geyer, Soumya Ghosh
| Challenge: | Large language models (LLMs) are being used for context-grounded tasks like summarizing meetings and answering doctors' questions. |
| Approach: | They propose a technique to provide explanations for context-grounded text generation by assigning scores to parts of the context to quantify their influence on the model output. |
| Outcome: | The proposed framework can provide more faithful explanations of generated output than available alternatives, including LLM self-explanations. |
AI Steerability 360: A Toolkit for Steering Large Language Models (2026.acl-demo)
Copied to clipboard
Erik Miehling, Karthikeyan Natesan Ramamurthy, Praveen Venkateswaran, Ching-Yun Ko, Pierre Dognin, Moninder Singh, Tejaswini Pedapati, Avinash Balakrishnan, Matthew Riemer, Dennis Wei, Inge Vejsbjerg, Elizabeth M. Daly, Kush R. Varshney
| Challenge: | The AI Steerability 360 toolkit is an extensible, open-source Python library for steering LLMs. |
| Approach: | The AI Steerability 360 toolkit is an extensible, open-source Python library for steering LLMs. |
| Outcome: | The toolkit is available under an Apache 2.0 license and is available on https://github.com/IBM/AISteer360. |
Your fairness may vary: Pretrained language model fairness in toxic text classification (2022.findings-acl)
Copied to clipboard
| Challenge: | Pre-trained, bidirectional language models have revolutionized natural language processing research . authors show that focusing on accuracy measures alone can lead to models with wide variation in fairness characteristics . |
| Approach: | They propose to use two post-processing methods to improve model fairness without retraining . they use pretrained language models of varying sizes on two toxic text classification tasks . |
| Outcome: | The proposed methods improve model fairness without retraining . the results show that the fairness variation is more than just accuracy . |